boreal forest disturbance and streamflow response ... and metcalfe 2000.pdfboreal forest disturbance...

Boreal forest disturbance and streamflowresponse, northeastern Ontario

J.M. Buttle and R.A. Metcalfe

Abstract: The effects of forest disturbance on streamflow from small (<10 km2) basins have been well documented;however, implications of such disturbance for streamflow from relatively large rivers in the Canadian boreal forest areunclear. Landsat imagery was used to determine changes in the type, amount, and location of forest disturbance innortheastern Ontario between 1985 and 1990. These were compared with streamflow responses from medium- andlarge-scale basins in the region. Harvesting dominated forest disturbance, and total disturbance as of 1990 ranged from25% of basin area in the northwest part of the region to 5% in the southeast. There was limited streamflow response toland cover changes, with no definitive changes in water year runoff or peak flow magnitude and timing. This likely re-flects the ability of relatively large basins to buffer the hydrologic impacts of the small degree of recent forest distur-bance, combined with the influence of climatic variability on temporal trends in basin streamflow. However,disturbance was associated with increases in moderate and low flows from medium and large basins, respectively,which occurred largely during summer months.

Résumé: Les effets des perturbations de la forêt sur l’écoulement dans les petits bassins (<10 km2) sont bien docu-mentés, mais les incidences de ces perturbations sur le débit de cours d’eau relativement grands dans la forêt boréalecanadienne ne sont pas clairs. Nous avons eu recours à l’imagerie Landsat pour déterminer les changements dans letype, l’importance et la localisation des perturbations de la forêt dans le nord-ouest de l’Ontario entre 1985 et 1990.Nous avons comparé ces éléments aux réactions dans l’écoulement de bassins d’échelle moyenne à grande dans la ré-gion. C’est l’exploitation forestière qui dominait les perturbations, et la perturbation totale en 1990 allait de 25% de lasuperficie du bassin dans la zone nord-ouest à 5% dans le sud-est. La réaction de l’écoulement au changement dans lacouverture du sol était limitée, et aucun changement définitif n’a été noté dans le volume d’écoulement annuel ni dansl’ampleur et le calendrier des pics de débit. Cela s’explique vraisemblablement par l’aptitude des bassins relativementgrands à compenser les impacts hydrologiques d’un faible degré de perturbation récente de la forêt, combinée àl’influence de la variabilité climatique sur les tendances temporelles du débit des bassins. Toutefois, la perturbationétait associée à des hausses des écoulements modérés et faibles des bassins de taille moyenne et grande, respective-ment, qui se produisaient essentiellement pendant les mois d’été.

[Traduit par la Rédaction] Buttle and Metcalfe 18

Introduction

The boreal forest is the dominant forest type in Canada,covering-3 million km2 (Zoltai et al. 1988). Frequency andintensity of disturbance of forest cover by fire are increasingin some parts of the boreal region (e.g., northwestern On-tario; Stocks and Street 1983) while decreasing elsewhere(e.g., northwestern Quebec; Dansereau and Bergeron 1993).Nevertheless, there appears to be a general increase in therate and magnitude of disturbance by harvesting in manyparts of the boreal forest (Schindler 1998). Research in theboreal forest, and to a greater extent in other forest types,has shown that forest disturbance can affect a range of as-pects of basin streamflow. Total runoff (water yield) gener-

ally increases with forest disturbance (Bosch and Hewlett1982; Sahin and Hall 1996), with reduced interception andtranspiration leading to greater streamflows. Total runoff in-creases roughly in proportion to the fraction of basin basalarea harvested (Keenan and Kimmins 1993), although theexact relationship varies between regions and forest type(Sahin and Hall 1996; Stednick 1996). The literature pro-vides “mixed messages about peak flow responses” (Thomasand Megahan 1998, p. 3402) to forest disturbance, with re-ports of increases (Verry et al. 1983; Jones and Grant 1996),increases for only small peak flows (Thomas and Megahan1998), decreases (Harr and McCorisin 1979), and no signifi-cant changes (Thomas and Megahan 1998) following har-vesting. Variations in peak flow response to disturbancelikely reflect intersite differences in such factors as the dom-inant runoff-generating process(es), climate, geology, topog-raphy, vegetation cover, soils, and the harvesting strategyemployed. There is greater consensus that summer rainfallsthat do not generate a streamflow response in forested basinsmay produce small to moderate flows in disturbed basins(Pomeroy et al. 1997). This is due to greater soil water stor-age resulting from reduced evaporation (Elliot et al. 1997),combined with soil compaction and enhanced potential forsoil saturation and overland flow. Disturbance may also alter

Can. J. Fish. Aquat. Sci.57(Suppl. 2): 5–18 (2000) © 2000 NRC Canada

5

Received August 10, 1999. Accepted May 4, 2000.J15298

J.M. Buttle.1 Department of Geography, Trent University,1600 West Bank Drive, Peterborough, ON K9J 7B8, Canada.R.A. Metcalfe. Department of Geography, Queen’sUniversity, 99 University Avenue, Kingston, ON K7L 3N6,Canada.

1Author to whom all correspondence should be addressed.e-mail: [email protected]

J:\cjfas\cjfas57\Supp2\F00-107.vpWednesday, August 30, 2000 7:17:16 AM

Color profile: DisabledComposite Default screen

the timing of peak flows. Snow cover in disturbed areas ab-lates before appreciable melt in forest stands. This producesseveral small peak flows, rather than a large peak flow frombasin-wide snowmelt (Verry et al. 1983). The date of the an-nual peak flow may also be advanced (Cheng 1989). Thereappears to be a general increase in low flows following dis-turbance due to wetting-up of the basin because of reduc-tions in rain and snow interception, sublimation, andevaporation (e.g., Hartman and Scrivener 1990; Dubé et al.1995; Prevost et al. 1999).

Most studies of the hydrological impacts of forest distur-bance have been conducted at small scales (<1 to 10 km2;Bosch and Hewlett 1982; Sachs et al. 1998), and the cumu-lative impacts of these local disturbances on streamflow atthe medium- (-100 km2) or large-basin (-1000 km2) scalesare unclear (Cheng 1989; Sahin and Hall 1996). Keenan andKimmins (1993) contended that if only a small proportion offirst-order basins experience increased streamflow due toharvesting, these effects are much reduced in second- andthird-order basins and are often undetectable in fourth- andfifth-order streams. Cheng (1989) found significant changesin streamflow from forested basins >10 km2 in area, pro-vided a sufficient portion of the basin was harvested. Coatsand Miller (1981) noted that the cumulative effects of con-centrated harvesting in large basins can lead to larger im-pacts on major rivers relative to those in smaller basins. Thevalidity of predicting the cumulative impact of forest distur-bance on large river systems by simply scaling up experi-mental results from small basins has been questioned(Cheng 1989), based on such concerns as uncertainties aboutflood arrival times from different tributaries and the inverserelationship between the frequency of runoff events of agiven magnitude and basin size (Coats and Miller 1981).

Despite the uncertainty regarding the cumulative effects offorest disturbance on the hydrology of medium- andlarge-scale basins, information is needed on the streamflowresponse to disturbance at these spatial scales in order to ad-dress the associated implications for such issues as aquaticecology, water supply, and generation of hydroelectricity.This paper examines (i) recent forest disturbance patterns inthe boreal forest of northeastern Ontario and (ii ) streamflowresponses from medium- and large-scale basins in the regionto this disturbance.

Materials and methods

Study areaThe study basins are the southwestern and southern tributaries

of the Moose River basin, which drains northeastern Ontario andpart of western Quebec northward to James Bay (Fig. 1; Table 1).They were selected to examine the effects of forest disturbance onstreamflow at the medium- (04LK001 and 04MD004, henceforthreferred to as M1 and M2) and large-basin (04LJ001, 04LF001,04LD001, and 04LB001, henceforth referred to as L1, L2, L3, andL4) scales. The basins overlie Precambrian Shield igneous andmetamorphic rocks (Superior Province) of Neo- to Meso-archeanage (2.5–3.4 Ga) and have limited relief and low gradients. Lowerportions of the medium basins overlie till plain, while higher eleva-tions are on lacustrine plain. Stream headwaters in the large basinsoriginate on Shield uplands and traverse the lacustrine plain to out-lets on the till plain.

Upland portions of the basins are largely mantled bywell-drained orthic humoferric podzols with a stony–rocky phase,while lower elevations are dominated by well-drained gray luvisolsand poorly drained mesisols and orthic gleysols (Clayton et al.1977a, 1977b). Forests (including treed wetlands) cover from 79(M2) to 89% (L2) of basin area. Forests in upland areas are domi-nated by black and white spruce, along with balsam fir, jack pine,white birch, and trembling aspen, while forested wetlands containtamarack and black spruce. Wetlands, cutovers, lakes, and agricul-tural, mining, and urban areas comprise a small proportion of eachbasin’s surface cover (Buttle et al. 1998).

The basins are in the Dfb modified Köppen climate zone: conti-nental cool summer boreal forest climate, adequate precipitationthroughout the year for vegetative growth, 4 or more months withmean temperatures >10°C, and mean temperatures in the coldestmonth lower than –3°C (Royal Commission on the Northern Envi-ronment 1985). There is a general increase in water year (WY: 1October of one year to 30 September of the next) precipitationfrom northwest to southeast across the study area, while the frac-tion of WY precipitation contributed by snowfall decreases alongthe same gradient (Table 2). Mean WY runoff ranges from-350 to-450 mm, while mean WY evaporation (residual of simplified wa-ter balance) ranges from-410 to-453 mm. Flow from L2, L3, andL4 is regulated as a result of hydroelectric dam operation. Regula-tion has reduced peak runoffs relative to unregulated basins, withthe greatest impact on flows equaled or exceeded <5% of the time(Buttle et al. 1998).

DisturbanceDisturbance was assessed using two classified Landsat images

© 2000 NRC Canada

6 Can. J. Fish. Aquat. Sci. Vol. 57(Suppl. 2), 2000

Physiographic units (fractionof basin area)

Water Survey ofCanada stationNo. (basin code) Station name

Area(km2)

Lowestelevation(m asl)

Highestelevation(m asl)

Reliefratioa

Tillplain

Lacustrineplain Upland

04LK001 (M1) Mattawishkwia River at Hearst 1 140 230 334 0.001 30 0.42 0.58 0.0004LJ001 (L1) Missinaibi River at Mattice 8 940 224 512 0.001 52 0.22 0.54 0.2504LF001 (L2) Kapuskasing River at Kapuskasing 6 760 201 500 0.001 39 0.31 0.19 0.5004LD001 (L3) Groundhog River at Fauquier 11 900 275 507 0.002 13 0.10 0.36 0.5404LB001 (L4) Mattagami River at Smooth Rock Falls 10 000 215 451 0.000 99 0.15 0.30 0.5504MD004 (M2) Porcupine River at Hoyle 401 271 310 0.001 26 0.06 0.94 0.00

aRelief ratio =H/L, whereH is difference between the highest and lowest points in the basin andL is the horizontal distance along the longestdimension of the basin parallel to the principal drainage line.

Table 1. Physical properties of the study basins.



from the Ontario Ministry of Natural Resources. The first was amultispectral scanner (MSS) image based on summer scenes at100-m resolution. Scenes were taken on various dates in 1984 and1985 to achieve cloud-free coverage of the Moose River basin.Classification employed supervised training and a maximum likeli-hood classifier on four bands of MSS data (White and Ellis 1987).The second image was a thematic mapper (TM) image based onsummer scenes taken at 25-m resolution in 1990 and 1991. TheTM classification was resampled to match the 100-m resolution ofthe MSS classification. Classification employed supervised train-ing and a maximum likelihood classifier on seven bands of TMdata (Spectranalysis Inc. 1997). Disturbance was classified as“cutovers” or “burns” in the MSS image and as “recent cutovers”(<10 years old), “recent burns” (<10 years old), and “old burns andcutovers” (up to 50 years old) in the TM image (Fig. 2).

To relate forest disturbance to basin streamflow, the MSS andTM images were assumed to represent conditions as of 1985 and1990, respectively. The 1985 classification does not indicate the

age of cutovers or burns, such that their state of regeneration whenthe imagery was taken is unknown. Therefore, cutover and burnclasses on the 1985 image represent the upper bound of disturbedarea, since some areas may have been in an advanced state of re-generation and may have been similar hydrologically to matureforest stands (cf. Pomeroy et al. 1997). There were no major firesin the basins during the 1976–1996 period according to data fromthe Fire Science and Technology Branch, Ontario Ministry of Nat-ural Resources, and areas classified in the 1985 image as burns andin the 1990 image as recent burns represented <0.01% of the com-bined basin areas. Therefore, burns and recent burns were not in-cluded in the analyses. Some 1985 cutovers were classified asrecent cutovers in the 1990 image. These areas would not havebeen harvested between 1985 and 1990 and were subtracted fromrecent cutovers in the 1990 image to estimate the extent of harvest-ing between 1985 and 1990. The accuracy of this assumption wasassessed by comparing the estimated extent of harvesting (squarekilometres) between 1985 and 1990 in an area in the west-central

© 2000 NRC Canada

Buttle and Metcalfe 7

Fig. 1. Location of the study basins. The heavily stippled area on the Ontario map indicates the area in the west-central portion of theMoose River basin for which estimated harvested area between 1985 and 1990 obtained from Landsat imagery was compared with ac-tual harvested area obtained from Ontario Ministry of Natural Resources records.



©2

00

0N

RC

Ca

na

da

8C

an.J.

Fish.A

quat.S

ci.Vol.

57(Suppl.

2),2000

Fig. 2. Disturbed areas in the study basins based on the 1985 and 1990 images. Burns (1985 image) and recent burns (1990 image) comprised a minor portion of the basin ar-eas and were not plotted.

J:\cjfas\cjfas57\Supp2\F00-107.vp

Wednesday, August 30, 2000 7:17:30 AM

Color profile: Disabled

Composite Default screen

©2

00

0N

RC

Ca

na

da

Buttle

andM

etcalfe9

Basin Water Survey of Canada station namePeriod ofrecord

WY precipitation(mm)

WY snowfall/WYprecipitation

WY runoff(mm) R:P ratioa

WY evaporation(mm)

M1 Mattawishkwia River at Hearst 1986–1993 844±76 0.35±0.06 417±103 0.49±0.10 427±75L1 Missinaibi River at Mattice 1920–1995 811±100 0.35±0.04 366±85 0.45±0.09 446±91L2 Kapuskasing River at Kapuskasing 1920–1995 807±98 0.33±0.04 360±78 0.45±0.08 446±89L3 Groundhog River at Fauquier 1920–1995 802±97 0.33±0.04 382±80 0.48±0.09 417±87L4 Mattagami River at Smooth Rock Falls 1920–1995 813±102 0.33±0.04 351±74 0.43±0.09 462±96M2 Porcupine River at Hoyle 1977–1993 895±77 0.31±0.04 442±68 0.49±0.06 453±63

aWY runoff ÷ WY precipitation.

Table 2. Water balance summary statistics for study basins (mean ± 1 SD) (from Buttle et al. 1998).

M1 L1 L2 L3 L4 M2

1985 1990 1985 1990 1985 1990 1985 1990 1985 1990 1985 1990

1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3

Area (km2) 49 7 209 922 603 1654 345 220 615 640 526 761 788 277 582 34 1 19

% of basin 4.3 0.6 18.3 10.7 6.7 18.5 5.0 3.3 9.1 5.1 4.4 6.4 8.1 2.8 5.8 8.5 0.2 4.7

Largest patch index (%)a 2.0 0.3 4.8 1.0 6.8 3.5 0.5 0.4 0.2 0.4 0.6 0.2 0.9 0.4 0.2 2.5 0.2 0.9

No. of patches 56 66 1184 810 4203 10 013 565 2000 5368 1031 3884 7263 810 2696 6106 59 20 202

Patch density (no.·km–2) 0.05 0.06 1.04 0.09 0.49 1.16 0.08 0.29 0.78 0.08 0.31 0.58 0.08 0.28 0.62 0.15 0.05 0.51

Mean patch size (km2) 0.92 0.12 0.19 1.02 0.25 0.14 0.64 0.15 0.10 0.64 0.19 0.08 1.02 0.16 0.07 0.58 0.08 0.09

CVN 0.141 0.045 1.140 0.352 0.510 0.989 0.204 0.238 0.608 0.213 0.288 0.391 0.278 0.257 0.393 0.421 0.025 0.359

Note: 1, cutovers; 2, areas estimated to have been harvested between 1985 and 1990; 3, old burns and cutovers on the 1990 TM image plus cutovers on the 1985 MSS image classed as recent cutovers on the1990 TM image.

aPercentage of basin area occupied by the largest patch.

Table 3. Disturbance characteristics of the study basins.

J:\cjfas\cjfas57\Supp2\F00-107.vp

Wednesday, August 30, 2000 7:17:31 AM

Color profile: Disabled

Composite Default screen

portion of the Moose River basin (Fig. 1) with the observed extentof harvesting between 1985 and 1990 in the same area based onOntario Ministry of Natural Resources data. Cells subtracted fromthe 1990 recent cutovers were added to the old burns and cutoversfrom the 1990 TM image, and total disturbance as of 1990 withineach basin was estimated as the sum of harvesting between 1985and 1990 plus the augmented old burns and cutovers.

The potential effect of forest disturbance on basin hydrology de-pends on several factors in addition to the amount of disturbancerelative to basin area. These include the nature of the surface onwhich disturbance occurred, the proximity of disturbed areas to thedrainage network (Thomas and Megahan 1998), and the pattern ofdisturbance within the landscape (i.e., fragmented or clumped).Relative differences in infiltration properties between surface typescould lead to variations in runoff response to the same amount ofharvesting on each surface type. Therefore, the extent of harvestingbetween 1985 and 1990 on each of the major physiographic unitsin the basins (Table 1) was determined. Harvesting of wetlandsgenerally increases water table levels (Dubé et al. 1995), and theconsequent expansion of saturated areas may enhance streamflowgeneration and increase peak flows and total runoff. However, theextent of disturbance on wetland surfaces was not assessed, sincethese localized impacts were felt to be undetectable at the scale ofbasins studied here.

In order to assess disturbance patterns with respect to the drain-age network, the Euclidean distance from each cell in a basin tothe nearest water body (major stream or lake) in the basin was cal-culated from 1:250 000 digital National Topographic Series drain-age network coverage. Cumulative total basin area and total area offorest disturbance with increasing distance from the drainage net-work were then determined for each basin. Using Euclidean dis-tance instead of flow pathways will tend to underestimate the truedistance of cells to the drainage network. A better estimate wouldbe obtained using drainage flowpaths, the accuracy of which is afunction of the basin digital topographic model used to estimateflowpath length. However, the size of the study area and resolutionof the available provincial digital topographic model (1 × 1 km)made derivation of flow pathways impractical.

Disturbance patterns were analysed for the 1985 and 1990 im-ages using indices at the patch and landscape scale. Each distur-bance type was analysed separately within each basin. Cellsclassified as one of the disturbance types were grouped into contig-uous patches if they contained the same attribute and touched inany of the eight possible cardinal directions. At the landscapescale, the contiguity of disturbed patches was assessed using thecenter versus neighbors (CVN) index (Murphy 1985). CVN wascalculated on a cell-by-cell basis using a 3 × 3 kernel, and themean was determined for each basin. The CVN is simply the num-ber of cells different from the center cell in the kernel. Index val-ues range from 0 (all cells of the same class) to 8 (center celldifferent from all neighbors), representing a decreasing degree ofcontiguity and an increasing degree of landscape fragmentation.

StreamflowThe paired-basin approach (with two or more basins of similar

geology, topography, soil cover, and climate) offers the best meansof assessing streamflow response to land-use change (Bosch andHewlett 1982) where one basin serves as a control, while landcover on the other basin(s) is modified. However, absence ofnearby control basins that did not experience land cover changesbetween 1985 and 1990 necessitated a “quasi-paired-basin ap-proach” (Buttle 1996), with partial control at a given scale pro-vided by the basin with the least forest disturbance during theperiod (i.e., M2 was the partial control for M1, and L4 was the par-tial control for L1, L2, and L3).

Water Survey of Canada daily mean discharges (cubic metresper second) for all basins for the period of record (Table 2) were

converted to daily runoff (millimetres oer day) to permit compari-son between basins of different sizes. WY runoff was calculatedfor each basin for WYs with complete records. Daily total precipi-tation and snowfall data were obtained from all Atmospheric Envi-ronment Service climate stations operating in and around thebasins for the period 1920–1996. WY total precipitation and totalsnowfall were calculated for each station by summing October–September monthly totals. These point data were then used to com-pute spatially averaged annual total precipitation and snowfall val-ues for each basin for any WYs with complete runoff records, asdescribed in Buttle et al. (1998). WY runoff was divided by totalprecipitation to obtain the WY runoff ratio.

Double-mass curves (DMCs) of cumulative runoff from M1 ver-sus M2 and from L1, L2, and L3 versus L4 were constructed. Lin-ear regression equations were fit to the medium-basin DMC.Missing streamflow data for L3 required division of the runoff re-cord into two periods: 1920–1966 and 1970–1993. Linear regres-sion equations were fit to the large-basin DMCs for each of thefollowing periods: 1920–1966, 1970–1993, 1970–1984, and1985–1993. The latter period approximated the time interval dur-ing which forest disturbance within the basins was assessed. DMCsof cumulative precipitation for M1 versus M2 and for L1, L2, andL3 versus L4 were also constructed for the same time periods usedfor the runoff DMCs.

Flow regulation of L2, L3, and L4 meant that the peak and lowflow response of only the unregulated basins (M1, M2, and L1)was examined for the period of concurrent flow record(1986–1993). Partial duration series were constructed, consistingof all peak flows equal to or greater than the smallest annual maxi-mum flow for the basin for the 1986–1993 period. Each peak flowwas categorized according to the type of generating event(snowmelt versus rainfall versus rain-on-snow) using AtmosphericEnvironment Service records from the nearest station. Separationof peak flows according to generating process has assisted interpre-tation of peak flow generation in both southern (Irvine and Drake1987) and northern Ontario (Woo and Waylen 1984) and examina-tion of peak flow responses to forest cover changes in southernOntario (Buttle 1994). An 8-day antecedent rainfall index (ARI)was calculated for the peak flow for each basin by summing allrain falling on the basin on the date of the peak flow and for the7 days preceding this date. An 8-day period was also used in Wooand Waylen’s (1984) analysis of floods in northern Ontario to ac-count for the significant time lag between water inputs (snowmeltor rainfall) and flood occurrence at the Water Survey of Canadastations resulting from the large basin size and ample wetland stor-age. Peaks were separated according to generating event assumingthat events between Julian days 1 and 140 and with 8-day ARI <25 mm were snowmelt events, events between Julian days 1 and140 and with 8-day ARI > 25 mm were rain-on-snow events, andall other events were rainfall events.

Flow duration curves (FDCs, cumulative frequency curvesshowing the fraction of time that specified daily runoffs wereequaled or exceeded for a period of record) were determined annu-ally for each of the unregulated basins for 1986–1993.

Results

Forest disturbanceThe 1990 TM image classified 1501 km2 of the area in the

west-central portion of the Moose River basin (Fig. 1) as re-cent cutovers. However, 559 km2 of these recent cutoverswere classified as cutovers in the 1985 MSS image. Removalof these areas from the 1990 recent cutovers gave an esti-mated harvested area of 942 km2 between 1985 and 1990.This was within-10% of the observed harvested area of1047 km2 during the same period. Thus, our approach ap-

© 2000 NRC Canada




pears to provide reasonable estimates of forest disturbancebetween 1985 and 1990 in the basins.

There were large differences in the amount and nature of

forest disturbance between basins for a given year and be-tween years for a given basin (Fig. 2; Table 3). Prior to1985, L1, M2, and L4 showed the greatest degree of forest

© 2000 NRC Canada


Fig. 3. (a) Percentage of basin area within a given distance of a major stream or lake, (b) percentage of basin area within a given dis-tance of a major stream or lake that had been disturbed as of 1990, (c) cumulative percentage of basin area that had been harvested asof 1985 within a given distance of a major stream or lake, (d) cumulative percentage of basin area that had been harvested between1985 and 1990 within a given distance of a major stream or lake, (e) percentage of basin area within a given distance of a majorstream or lake that had been harvested as of 1985, and (f) percentage of basin area within a given distance of a major stream or lakethat had been harvested between 1985 and 1990.



disturbance. However, only 0.2% of M2 was harvested be-tween 1985 and 1990, while 6.7% of L1 was harvested dur-ing the same period. Total disturbance (harvesting between1985 and 1990 plus old burns and cutovers on the 1990 TMimage plus cutovers from the 1985 MSS image classed asrecent cutovers on the 1990 TM image) generally declinedfrom west to east across the study area: M1 (18.9%), L1

(25.2%), L2 (12.4%), L3 (10.8%), L4 (8.6%), and M2(4.9%).

Harvesting intensity on major physiographic units variedbetween basins. For large basins, the proportion of harvest-ing between 1985 and 1990 on upland areas was 20 (L1), 67(L2), 57 (L3), and 70% (L4). Most harvesting in L1 (61%)occurred on lacustrine plain. Of the medium basins, 90% of

© 2000 NRC Canada


Fig. 4. DMCs of cumulative runoff: (a) M1 versus M2 (1986–1993) and linear best fits to the DMC for the 1 January 1986 – 1 May1992 (thick short-dashed line,Y = 0.845X – 0.048,r 2 = 0.997) and 1 June 1992 – 14 September 1993 (thick long-dashed line,Y =1.038X – 0.385,r 2 = 0.983) periods and L1 (thick long-dashed line), L2 (thick short-dashed line), and L3 (solid line) versus L4 forthe (b) 1920–1966 and (c) 1970–1993 periods.



harvesting between 1985 and 1990 occurred on till plain and10% on lacustrine plain in M1. This pattern was reversed inM2, where 98% of harvesting was on lacustrine plain andonly 2% on till plain.

The proportion of basin area within 1 km of majorstreams or lakes was greater for large compared with me-dium basins (Fig. 3a), suggesting that harvesting in largebasins will also tend to be relatively closer to major waterbodies. This was supported by the 1985 Landsat image,where the fraction of total harvesting within 1 km of majorwater bodies (Fig. 3c) ranged from 0.77 to 0.85 for largebasins but was only 0.7 and 0.72 in M1 and M2, respec-tively. The fraction of total harvesting between 1985 and

1990 within 1 km of major streams and lakes was similar to1985 results for large basins (Fig. 3d); however, it increasedto 0.84 for M1 while declining to 0.29 for M2.

Distribution of harvesting intensity within each basin wasdetermined by dividing the amount of harvesting within agiven distance interval from major water bodies by the basinarea within that distance interval (Figs. 3e and 3f). In almostall cases, harvesting intensity of basin area within100–200 m of major streams and lakes was greater than thatwithin 0–100 m. This may reflect the presence of bufferstrips, which can extend 30–90 m from streams or lakes (R.Mackereth, Centre for Northern Forest Ecosystem Research,Thunder Bay, Ont., personal communication) and in which

© 2000 NRC Canada


Fig. 5. Daily mean runoffs for M1, L1, and M2 (1986–1993) and type of event (diamonds, snowmelt; squares, rain-on-snow; circles,rainfall) that generated each peak flow in the partial duration series.



harvesting is generally not permitted. Some basins showedconsistent patterns between the two images; for example,harvesting intensity in L1 increased with distance from wa-ter bodies up to-2.5 km and dropped markedly beyond thatdistance. In other cases, the degree of harvesting intensitywith distance from major water bodies changed with time.Thus, the greatest harvesting intensity in L2 as of 1985 wason areas-500 m from streams and lakes, with disturbancedecreasing with distance. The 1990 image showed the re-verse pattern, with harvesting intensity increasing with dis-tance from streams or lakes and peaking at-2.2 km. Despitethis shift, distribution of total disturbance as of 1990(Fig. 3b) in L2 was relatively uniform with distance frommajor water bodies. Harvesting between 1985 and 1990 inM1 and M2 was minor (Fig. 3d). However, intensity of totaldisturbance was much greater in M1 (Fig. 3b), where >15%of the basin within 1 km of major water bodies was dis-turbed by 1990, most of this in the form of old burns andcutovers.

The nature of harvesting within basins changed with time

(Table 3). Mean patch size of harvested areas decreasedfrom 1985 to 1990 in all basins, while number of patchesand patch density increased in all but M2. The latter is con-sistent with the decrease in harvesting in this basin duringthe -5-year period. Patches comprising 1990 old burns andcutovers plus 1985 cutovers classed as 1990 recent cutoversgenerally showed the least patch contiguity in any given ba-sin. The exception was M2, where the 1985 cutover patcheswere the least contiguous of any disturbed areas in the basin.The 1990 image indicated that patches harvested between1985 and 1990 in L1 were the least contiguous of any of thebasins, with 1985 cutover patch contiguity being greatest inM1 and M2. However, the range in CVN values for1985–1990 harvested patches between basins was small(0.025–0.510) relative to the potential range (0–8).

Streamflow response

RunoffThere was good agreement between cumulative daily run-

© 2000 NRC Canada


Time period L1 vs. L4 L2 vs. L4 L3 vs. L4

1920–1966 1.051a (0.00%) 1.042a (0.00%) 1.074a (0.00%)1970–1993 1.094a,b (4.09%) 1.027a,c (–1.44%) 1.143a,b (6.42%)1970–1984 1.086a,b (3.33%) 1.014 (–2.69%) 1.115a,b (3.82%)1985–1993 1.177a,b,d (11.99%) 1.103a,b,d (5.85%) 1.201a,b,d (10.57%)

Note: Coefficients of explanation (r 2) for all equations were >0.99. Values in parentheses indicate percentchange in slope coefficient relative to the coefficient for the 1920–1966 period, i.e., ((coefficient for periodx –coefficient for 1920–1966)/coefficient for 1920–1966) × 100%.

aSignificantly greater than 1 (p < 0.05).bSignificantly greater than slope coefficient for the 1920–1966 period (p < 0.05).cSignificantly less than slope coefficient for the 1920–1966 period (p < 0.05).dSignificantly greater than slope coefficient for the 1970–1984 period (p < 0.05).

Table 4. Slope coefficients for linear regression equations fit to DMCs relating cumulative WYrunoff from L1, L2, and L3 to cumulative runoff from L4.

% of time daily mean runoff equaled or exceeded

Basin Year 1 5 10 25 50 75 100

M2 1986 9 5 2.9 1.15 0.58 0.305 0.171987 3.4 2.4 1.95 1.1 0.7 0.31 0.171988 11 5.05 4 1.7 0.7 0.28 0.111989 9.2 6 3.9 1.05 0.5 0.28 0.191990 11.5 5.1 3.3 1.9 0.9 0.28 0.1081991 8.3 5.1 1.7 0.74 0.33 0.2 0.0761992 10.5 5.9 2 1.05 0.41 0.18 0.084

M1 1986 6.6 4.1 1.8 0.79 0.19 0.072 0.031987 6.3 3.4 2.4 1.35 0.53 0.08 0.0311988 8.1 5.7 3.2 1.5 0.59 0.14 0.031989 9 7 2.05 0.53 0.14 0.065 0.0131990 16 4.6 3.05 1.55 0.57 0.16 0.0391991 6.05 4.3 3 1.4 0.24 0.065 0.0041992 19 5.9 3 1.7 0.69 0.105 0.037

L1 1986 5.4 3.7 2 0.695 0.37 0.17 0.0831987 3.5 2 1.15 0.73 0.47 0.2 0.1051988 5 3.35 2.75 1.3 0.59 0.205 0.1051989 8.05 6 2.6 0.975 0.325 0.2 0.1051990 10 4.05 2.5 1.6 0.69 0.24 0.111991 5.4 3.6 2.3 1.05 0.37 0.2 0.0731992 8.9 7.5 2.1 1.2 0.67 0.3 0.115

Table 5. Daily mean runoffs (mm·day–1) corresponding to various flow durations, 1986–1992.



off from the M1 and M2 basins from 1 January 1986 to 1May 1992 (Fig. 4a), with cumulative runoff from M1 aver-aging-85% of that from M2. This relationship changed fol-lowing a large event beginning 1 May 1992 in M1 (Fig. 5),when M2 generated comparatively less runoff during this31-day period. The slope of the DMC increased significantlyfollowing 1 June 1992, such that cumulative runoff from M1was-4% higher than that from M2. Conversely, the DMCof cumulative precipitation for M1 versus M2 (not shown)remained linear during the 1986–1992 period.

Shifts in DMC relationships for large basins between peri-ods (Figs. 4b and 4c; Table 4) may have resulted from sys-tematic temporal trends in runoff from L4 due to climaticvariations and (or) land-use changes. However, time seriesanalysis of WY runoff from L4 showed no significant tem-poral trend during either the 1920–1966 or 1970–1993 pe-riod, which supported use of L4 as a “quasi-control” basin.DMCs of cumulative WY precipitation (not shown) indi-cated that precipitation in L4 generally equaled or exceededthat in L1, L2, and L3 for both the 1920–1966 and1970–1993 periods. Cumulative runoff from L3 exceededthat from other basins during each period. The reason forthis is unknown, although it is consistent with L3’s highermean runoff/precipitation ratio relative to the other basins(Table 2).

Slope coefficients of linear best fits to DMCs from all bas-ins were significantly >1 for all periods, with the exceptionof the slope coefficient for L2 versus L4 for the 1970–1984period. Thus runoff from L1, L2, and L3 generally exceededthat from L4. This difference was most pronounced for the1985–1993 period, and DMC slopes ranged from 5.85 (L2)to 11.99% (L1) greater than the slope of the correspondingcurve for the 1920–1966 period. The increase in WY runoffrelative to that expected from the 1920–1966 period can bepredicted using the 1985–1993 DMC slopes, assuming amean WY runoff of 351 mm for L4 (Table 2). Thus, WYrunoff would be 44 (L1), 21 (L2), and 45 mm (L3) greaterthan that estimated by the DMC slopes for the 1920–1966period.

WY runoff ratiosThere were no significant temporal trends in the ratio of

WY runoff to WY precipitation for medium or large basins,either for original ratios or for differences in ratios betweenbasins. Total precipitation, total snowfall, or the fraction ofWY precipitation consisting of snowfall did not explain asignificant amount of the variation in runoff ratios for eachbasin.

Peak flowsAnnual maxima in M2 were generated by snowmelt in

1986, 1987, 1988, 1989, and 1991 and by rain-on-snow inthe remaining years (Fig. 5). Annual maxima at the start ofthe period were similar to those at the end, with no apparenttemporal trend. Rainfall generated a larger number of peakflows for M2 than for M1 and L1. Most annual maxima inM1 were generated by snowmelt. The major exception wasthe rainfall-generated annual maximum in 1993, whilerain-on-snow peak flows occurred only in 1988. Snowmeltin 1990 and 1992 generated annual maxima approximatelytwice those observed at the beginning of the period; how-ever, there was no statistically significant temporal trend in

the snowmelt peak flows. L1 showed a similar pattern tothat of M1. Annual maxima were generated by snowmelt inalmost all years, while rain-on-snow resulted in only one an-nual maximum flow (1988). Snowmelt in L1 also tended togenerate larger peak flows later in the period relative tothose seen in 1986, 1987, and 1988; however, no statisticallysignificant trend was observed with time.

FDCs and timing of annual minimum flowsDaily mean runoffs corresponding to various flow

exceedance levels were abstracted from the annual FDCs forM1, L1, and M2 (Table 5). Runoff at a given exceedancelevel for a particular year was expressed relative to runoff atthe same exceedance level in 1986:

(1) RatioRunoff runoff

Runoffyear

yearx

x=

- 1986

1986

Ratios were determined for the 100, 75, 50, 25, 10, 5, and1% exceedance levels. The mean ratio for M2 for all yearswas not significantly different from 0 for any flowexceedance level, and there were no statistically significanttemporal trends in any ratios. Temporal trends in ratios forboth M1 and L1 were also statistically insignificant. Meanratios at the largest and smallest flow exceedance levels (1,5, 75, and 100%) in M1 were not significantly different from0 (difference of meanst test); however, mean ratios weresignificantly >0 at the 10, 25, and 50% levels. This indicatesthat there had been in increase in moderate flows in the ba-sin during the 1986–1992 period. Mean ratios at the 1, 5, 10,and 50% exceedance levels in L1 were not significantly dif-ferent from 0. However, mean ratios were significantly >0 atthe 25, 75, and 100% levels, indicating an increase in mod-erate and low flows during the 1986–1992 period.

Annual minima between 1986 and 1993 occurred either inmid- to late summer or in late winter, immediately beforethe onset of spring snowmelt. There was good agreement(within 1 or 2 days) between the dates of occurrence of an-nual minimum flow for M1 and L1. This is not surprisinggiven the proximity of the basins. The date of annual mini-mum flow for these two basins and M2 agreed to within afew days in four out of eight years. There was no systematicdifference between the two western basins and M2 in the re-maining years. Annual minima occurred in late winter in M1and L1 and in mid- to late summer in M2 in some years,while the reverse occurred in other years. There was also nosignificant correlation between annual variations in magni-tude of minimum flows from the basins between 1986 and1993.

Discussion

Most of the recent forest disturbance in the boreal forestof northeastern Ontario has resulted from harvesting, whilefire (either natural or as prescribed burns) was of minimalsignificance at the spatial scale examined here. Harvestingpatterns have changed in recent years, with mean cutblocksize decreasing from 1985 to 1990. Southeastern basins inthe study area experienced either no change or a decrease inharvesting activity (as a percentage of basin area) during the1985–1990 period. Conversely, harvesting increased in otherbasins during this period, particularly in the northwest part

© 2000 NRC Canada




of the study area. The small range in CVN values for1985–1990 harvested patches between basins suggests thatrecent harvesting in the study area involved small but rela-tively contiguous cutblocks that have produced a low degreeof landscape fragmentation. Sachs et al. (1998) reported asimilar situation in interior British Columbia, where thelandscape is in the early stages of fragmentation as a resultof harvesting.

Assessment of the impacts of forest disturbance onstreamflow in this region is complicated by the large basinsizes, flow regulation and limited streamflow records forsome basins, lack of proper control basins, limited extent offorest disturbance, and availability of data on forest distur-bance for only two instants in time. Thus, we did not ob-served significant changes in such properties as runoff/precipitation ratios that have been noted for small basinselsewhere in the boreal forest (e.g., Bayley et al. 1992).

Nevertheless, recent forest disturbance appears to be asso-ciated with some aspects of streamflow from medium andlarge basins in the region. Increased runoff from M1 relativeto M2 following 1 May 1992 reflects the greater proportionof M1 that was disturbed as of 1990 (18.9 versus 4.9%). Theshift in the runoff DMC for M1 versus M2 was likely not theresult of changes in relative precipitation inputs to the bas-ins, which were fairly constant during the1986–1992 period.However, this change cannot be attributed to forest distur-bance with great confidence, due to brevity of the hydrologi-cal record and lack of data on temporal changes in forestdisturbance in M1 during the 1985–1990 period.

Equivocal links between forest disturbance and runoffwere also apparent for the large basins. L1, L2, and L3 alldemonstrated increased runoff relative to L4, particularlyduring the 1985–1995 period. However, these basins did notexperience increased precipitation inputs relative to those forL4. Total disturbed area as of 1990 in L4 was 8.6% of basinarea. Disturbance in the other large basins in excess of this“quasi-control” value was 2.2 (L3), 3.8 (L2), and 16.6%(L1). These forest cover changes are below the 20–25%threshold that must be met or exceeded to detect a measur-able response in annual runoff to forest disturbance (Boschand Hewlett 1982; Hornbeck et al. 1993). Nevertheless,Trimble et al. (1987) and Buttle (1994) documented signifi-cant changes in annual runoff for changes in forest cover be-low this threshold. The increase in DMC slope for L1 isconsistent with total forest disturbance, concentration of har-vesting in the basin on the lacustrine plain with its compara-tively low infiltration rates, mean cutblock patch size, andrecent harvesting intensity in near-stream areas in L1 rela-tive to the other large basins. The 44-mm increase in WYrunoff is also bracketed by a 23-mm increase predicted for a10% reduction in coniferous forest cover (Sahin and Hall1996) and by a 100-mm increase in annual runoff followingclear-cutting of small (<4.5 km2) boreal forest basins innorth-central Ontario (Nicolson 1995). However, the slightdifferences in 1985–1990 harvesting activity and total distur-bance between L2, L3, and L4 appear insufficient to accountfor the increased cumulative runoff for L2 and L3 relative toL4 during the 1985–1993 period. Increased runoff from L2and L3 was also unrelated to distribution of recent harvest-ing or total disturbance with respect to major water bodies,which was similar for L2, L3, and L4. Despite the results of

time series analysis of WY runoff from L4, climate changemay have influenced the runoff response of the large basins.Thus, the increased WY runoff from L1 cannot be attributedto forest disturbance in the basin. Paterson et al. (1998) sug-gested that hydrological changes induced by climatic varia-tions in the boreal forest may override those due to forestdisturbance such as harvesting or fire for small basins, andthis should be examined in future work at larger spatialscales.

Forest disturbance did not produce significant changes inpeak flows in this region, regardless of the nature of the gen-erating event. Cheng (1989) found increases in spring runofffollowing clear-cutting in interior British Columbia to be re-lated to the amount of rainfall and snowmelt during Apriland May (Cheng 1989). We do not have information onsnowmelt rates in each basin during the 1986–1993 period,and no relationship was found between the size ofsnowmelt-generated peak flows and snowfall during the pre-ceding winter. Nevertheless, the similarity in annual maximagenerated by snowmelt and rain-on-snow between 1986 and1993 for M2 suggests no significant interannual changes inthe intensity of snowpack ablation during this period. It maybe reasonable to assume that the meteorological conditionsthat drive snowmelt and that are relatively independent offorest cover conditions (i.e., incoming solar radiation, rain-fall, air mass wind speed, temperature, and relative humid-ity) are relatively similar between the basins in a given year.Thus, there is some association between the degree of forestdisturbance and increases in the size of snowmelt peak flowsfor some years in M1 and L1. However, the change in peakflows is not as conclusive as demonstrated elsewhere (e.g.,Verry et al. 1983; Cheng 1989), which may reflect the largesize of the study basins combined with the relatively smallfraction of the basins that were disturbed between 1985 and1990. Rapid snowmelt runoff from disturbed areas may sim-ply have been subsumed by the hydrological response of thelarge undisturbed portion of the basins.

All three basins had multiple flood peaks during snowmeltin most years, which might reflect the disturbance that allbasins had undergone prior to 1986. M2 had multiple floodpeaks in all but one snowmelt period, thus providing theclearest demonstration of desynchronization of flood peaks(cf. Verry et al. 1983). In addition to disturbance by harvest-ing, roughly 9% of M2 was impacted by urban development,agriculture, and mining as of 1990 (Buttle et al. 1998).Snowmelt in these open areas, superimposed on theclay-rich soils that dominate the basin, appeared to producea streamflow peak that preceded that due to melt in forestedparts of M2. Despite forest disturbance in M1 and L1 be-tween 1985 and 1990, we found no evidence for increasingdesynchronization of flood peaks during snowmelt. Therewas also no advance in the date of the initial snowmelt peakflow from any of the basins between 1986 and 1993. Aswith peak flow magnitude, the basins may have been toolarge to detect the effect of disturbance on peak flow timing.

The FDC analyses suggest that the minor forest distur-bance in M2 between 1985 and 1990 did not result in signif-icant changes in streamflow. As with the partial durationseries, the FDCs indicate that while peak flows for M1 andL1 had increased in some years during the mid- to latter partof the 1986–1993 period, there were no significant temporal

© 2000 NRC Canada




trends. Timing of low flows was also independent of the de-gree of basin disturbance. Nevertheless, forest disturbance inM1 and L1 was associated with increases in small and mod-erate flows. This is consistent with studies (e.g., Hornbeck etal. 1993; Keenan and Kimmins 1993; Hartman et al. 1996)showing clear-cutting to produce the greatest streamflow in-creases during low-flow periods, with smaller impacts onpeak flows. The results also support the scale effect in therelationship between forest disturbance and resulting stream-flow changes suggested by Coats and Miller (1981). Thus,disturbance was associated with increased moderate flows inthe medium basin but mainly with low flow increases in thelarge basin. These increases generally occurred during thesummer, and the associated changes in flow levels may haveimportant implications for aquatic communities in theserivers. This scale dependence of the hydrological response todisturbance requires further study and should be consideredwhen assessing the cumulative impact of forest disturbanceon the hydroecology of large river systems.

Acknowledgements

This work was supported by a contract from the OntarioMinistry of Natural Resources and by the Natural Sciencesand Engineering Research Council of Canada. We wish tothank Anthony Story and Hannah Wilson (Trent University)and Simon Bridge and Dave White (Ontario Ministry ofNatural Resources) for their assistance. Richard Carignan,André Plamondon, and an anonymous reviewer are thankedfor comments on a previous version of this paper.

References

Bayley, S.E., Schindler, D.W., Beaty, K.G., Parker, B.R., and Stainton,M.P. 1992. Effects of multiple fires on nutrient yields from streamsdraining boreal forest and fen watersheds: nitrogen and phospho-rus. Can. J. Fish. Aquat. Sci.49: 584–596.

Bosch, J.M., and Hewlett, J.D. 1982. A review of catchment exper-iments to determine the effect of vegetation changes on water,yield and evaporation. J. Hydrol.55: 3–23.

Buttle, J.M. 1994. Hydrological response to reforestation in the Gana-raska River basin, southern Ontario. Can. Geogr.38: 240–253.

Buttle, J.M. 1996. Identifying hydrological responses to basin restora-tion: an example from southern Ontario.In Watershed restorationmanagement: physical, chemical, and biological considerations.Edited byJ.J. McDonnell, J.B. Stribling, L.R. Neville, and D.J.Leopold. American Water Resources Association, Herndon, Va.pp. 5–13.

Buttle, J.M., Metcalfe, R.A., Story, A., and Wilson, H. 1998. Hydro-logic conditions, forest management and forest disturbance his-tory of the Moose River basin. A Report to the Moose RiverBasin Environmental Information Partnership, Northeast Region,Ontario Ministry of Natural Resources, South Porcupine, Ont.

Cheng, J.D. 1989. Streamflow changes after clear-cut logging of apine beetle infested watershed in southern British Columbia,Canada. Water. Resour. Res.25: 449–456.

Clayton, J.S., Ehrlich, W.A., Cann, D.B., Day, J.H., and Marshall,I.B. 1977a. Soils of Canada. Vol. 1. Soil report. AgricultureCanada, Ottawa, Ont.

Clayton, J.S., Ehrlich, W.A., Cann, D.B., Day, J.H., and Marshall,I.B. 1977b. Soils of Canada. Vol. 2. Inventory. Agriculture Can-ada, Ottawa, Ont.

Coats, R.N., and Miller, T.O. 1981. Cumulative silvicultural im-pacts on watersheds: a hydrological and regulatory dilemma.Environ. Manage.5: 147–160.

Dansereau, P-R., and Bergeron, Y. 1993. Fire history in the south-ern boreal forest of northwestern Quebec. Can. J. For. Res.23:25–32.

Dubé, S., Plamondon, A.P., and Rothwell, R.L. 1995. Watering-upafter clear-cutting on forested wetlands of the St. LawrenceLowlands. Water Resour. Res.31: 1741–1750.

Elliot, J.A., Toth, B.M., Granger, R.J., and Pomeroy, J.W. 1997.Soil moisture storage in mature and replanted sub-humid borealforest stands. Can. J. Soil Sci.78: 17–27.

Harr, R.D., and McCorisin, F.M. 1979. Initial effects of clearcutlogging on size and timing of peak flows in a small watershed inwestern Oregon. Water Resour. Res.15: 90–94.

Hartman, G.F., and Scrivener, J.C. 1990. Impacts of forestry prac-tices on a coastal stream ecosystem, Carnation Creek, BritishColumbia. Can. Bull. Fish. Aquat. Sci. No. 223.

Hartman, G.F., Scrivener, J.C., and Miles, M.J. 1996. Impacts oflogging in Carnation Creek, a high-energy coastal stream inBritish Columbia, and their implication for restoring fish habitat.Can. J. Fish. Aquat. Sci.53(Suppl. 1): 237–251.

Hornbeck, J.W., Adams, M.B., Corbett, E.S., Verry, E.S., and Lynch,J.A. 1993. Long-term impacts of forest treatments on water yield:a summary for northeastern U.S.A. J. Hydrol.150: 323–340.

Irvine, K.N., and Drake, J.J. 1987. Spatial analysis of snow- and rain-generated highflows in southern Ontario. Can. Geogr.31: 140–149.

Jones, J.A., and Grant, G.E. 1996. Peak flow responses toclear-cutting and roads in small and large basin, western Cas-cades, Oregon. Water Resour. Res.32: 959–974.

Keenan, R.J., and Kimmins, J.P. 1993. The ecological effects ofclear-cutting. Environ. Rev.1: 121–144.

Murphy, D.L. 1985. Estimating neighborhood variability with a binarycomparison matrix. Photogram. Eng. Remote Sens.51: 667–674.

Nicolson, J.A. 1995. Water quality and quantity in black sprucestands following harvesting. Tech. Note No. 74. Can. Forest Ser-vice, Sault Ste. Marie, Ont.

Paterson, A.M., Cumming, B.F, Smol, J.P., Blais, J.M., and France,R.L. 1998. Assessment of the effects of logging, forest fires, anddrought on lakes in northwestern Ontario: a 30-year paleo-limnological perspective. Can. J. For. Res.28: 1546–1556.

Pomeroy, J.W., Granger, R.J., Pietroniro, A., Elliott, J.E., Toth, B.,and Hedstrom, N. 1997. Hydrological pathways in the PrinceAlbert Model Forest. Final report to the Prince Albert ModelForest Association. National Hydrology Research Institute, En-vironment Canada, Saskatoon, Sask.

Prevost, M., Plamondon, A.P., and Belleau, P. 1999. Effects ofdrainage of a forested peatland on water quality and quantity. J.Hydrol. 214: 130–143.

Royal Commission on the Northern Environment. 1985. North of50°: an atlas of far northern Ontario. University of TorontoPress, Toronto, Ont.

Sachs, D.L., Sollins, P., and Cohen, W.B. 1998. Detecting land-scape changes in the interior of British Columbia from 1975 to1992 using satellite imagery. Can. J. For. Res.28: 23–36.

Sahin, V., and Hall, M.J. 1996. The effects of afforestation and de-forestation on water yields. J. Hydrol.178: 293–309.

Schindler, D.W. 1998. A dim future for boreal waters and land-scapes. BioScience,48: 157–164.

Spectranalysis Inc. 1997. User manual on the Ontario ProvincialLand Cover Data Base. Prepared for Natural Resources Manage-ment Information Branch, Ontario Ministry of Natural Resources,Oakville, Ont.

© 2000 NRC Canada




Stednick, J.D. 1996. Monitoring the effect of timber harvest on an-nual water yield. J. Hydrol.176: 79–95.

Stocks, B.J., and Street, R.B. 1983. Forest fire weather and wildfireoccurrence in the boreal forest of northwestern Ontario.In Re-sources and dynamics of the boreal zone.Edited byR.W. Wein,R.R. Riewe, and I.R. Methven. Association of Canadian Univer-sities for Northern Studies, Ottawa, Ont. pp. 249–265.

Thomas, R.B., and Megahan, W.F. 1998. Peak flow responses toclear-cutting and roads in small and large basins, western Cas-cades, Oregon: a second opinion. Water Resour. Res.34:3393–3403.

Trimble, S.W., Weirich, F.H., and Hoag, B.L. 1987. Reforestationand the reduction of water yield on the Southern Piedmont sincecirca 1940. Water Resour. Res.23: 425–437.

Verry, E.S., Lewis, J.R., and Brooks, K.N. 1983. Aspen clear-

cutting increases snowmelt and storm flow peaks in north cen-tral Minnesota. Water Resour. Bull.19: 59–67.

White, D.B., and Ellis, T.J. 1987. Providing input and output conti-nuity in Landsat-based digital mapping of land cover over largeareas.In Proceedings of the 11th Canadian Symposium on Re-mote Sensing, University of Waterloo, Waterloo, Ont. Departmentof Geography, University of Waterloo, Waterloo, Ont. pp. 81–85.

Woo, M-K, and Waylen, P.R. 1984. Areal prediction of annualfloods generated by two distinct processes. Hydrol. Sci. J.29:75–88.

Zoltai, S.C., Taylor, S., Jeglum, J.K., Mills, G.F., and Johnson, J.D.1988. Wetlands of boreal Canada.In Wetlands of Canada. Eco-logical Land Classification Series No. 24. Environment Canada,Ottawa, Ont., and Polyscience Publications, Inc., Montreal, Que.pp. 97–154.

© 2000 NRC Canada




boreal forest disturbance and streamflow response ... and metcalfe 2000.pdfboreal forest disturbance...

Documents